Add RXCROPMATURITY test to aux_clm

[samsrabin]

As discussed earlier with @ekluzek, I think it'd be good to have the RXCROPMATURITY system test run as part of the aux_clm test suite. However, in its usual configuration—SMS_Ly5 on the 10x15 global grid—it needs a walltime of 3 hours, since it runs the simulation twice and does some processing in between. This is as long as the longest existing test.

How short would I need to make it for it to be palatable as part of aux_clm?

Some ideas for shortening it:

• Reducing from 5 years to 4. This would mean that only one year is actually checked in my Python scripts, which is probably fine.
• Using a custom 10x15 grid that only includes every other latitude row. There are some edge cases that happen at high latitudes, and I also want to be sure to test tropical CFTs in their proper zone. (So I can't just say "high latitudes" or "tropics.")
• Using the PPE grid?

[billsacks]

Take a look at #275 (comment) and some of the earlier comments in that issue for what I have worked to accomplish recently in terms of total test time and cost. I don't feel like there's any particular hard cutoff, but keeping the test suite short-ish and cheap-ish is a good goal.

[samsrabin]

Marking as low-priority because there is at least a partial RXCROPMATURITY test in aux_clm now.

[samsrabin]

The "partial RXCROPMATURITY test in aux_clm" I referred to above is the RXCROPMATURITYSKIPGEN test. Unfortunately that test doesn't actually exercise the generate_gdds script, which is kinda the whole point.

I'm going to resolve this soon by instead making the RXCROPMATURITYSKIPGEN test run generate_gdds on archived outputs from a full RXCROPMATURITY test's GDD-generating run. I will also reduce the RXCROPMATURITYSKIPGEN test in the test list to run only 367 days instead of 61 months. This will make it quite painless to include in aux_clm.

Finally, I will rename RXCROPMATURITYSKIPGEN to RXCROPMATURITYSKIPFIRSTRUN to reflect the new behavior.